Optoelectronic Implementation of Progressive Deadlock Recovery-based Routing:

Timothy Mark Pinkston

University of Southern California


Résumé:

The development of fully-adaptive, cut-through networks with high bandwith is important for achieving scalable performance in communication-critical parallel processor systems. The router is the basic switching component that establishes the network topology and determines the allowable paths that packets can take through the network. Increased flexibility in routing allows available network bandwidth to be used efficiently but, also, creates more opportunity for cyclic resource dependencies to form which can result in deadlock. If not guarded against, deadlocks in routing can make packets block in the network indefinitely and, eventually, could result in the entire network coming to a complete standstill. This talk presents the design of a simple, flexible, and efficient optoelectronic router (the WARRP router) which is based on progressive deadlock recovery as opposed to deadlock avoidance or regressive deadlock recovery. Performance is optimized by allowing the maximum routing freedom provided by network resources to be exploited, which results in as much as 45% and 25% increased throughput over the best avoidance-based and regressive recovery-based routing techniques, respectively. Moreover, the pin-out problem is addressed by integrating dense high-speed optoelectronic I/O onto the CMOS-VLSI router core. Thus, this work also demonstrates the potential applicability of optoelectronic technology to next generation network router architectures that incorporate advanced routing techniques and require high bandwidth connectivity.


[Timothy Mark Pinkston]
[University of Southern California]